-
Notifications
You must be signed in to change notification settings - Fork 0
/
test_single.py
149 lines (123 loc) · 6.02 KB
/
test_single.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
import os
from glob import glob
from itertools import product
from time import perf_counter
from losses.cable import CableBSplineLoss
from losses.cable_pts import CableBSplinePtsLoss, CablePtsLoss
from models.cnn import CNN
from models.inbilstm import INBiLSTM
from models.separated_cnn_neural_predictor import SeparatedCNNNeuralPredictor
from models.separated_neural_predictor import SeparatedNeuralPredictor
from utils.bspline import BSpline
from utils.constants import BSplineConstants
from utils.geometry import calculateL3
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
from utils.dataset import _ds, prepare_dataset, whiten, mix_datasets, whitening, compute_ds_stats, unpack_translation, \
unpack_rotation, prepare_dataset_cond, unpack_cable
from utils.execution import ExperimentHandler
from models.basic_neural_predictor import BasicNeuralPredictor
np.random.seed(444)
#path = f"./trained_models/all_mb_03_27/new_mb_03_27_poc64_lr5em4_bs128_sep_diff_rotmat_dcable_augwithzeros_/checkpoints"
#path = f"./trained_models/all_mb_03_27/new_mb_03_27_poc64_lr5em4_bs128_sep_nodiff_rotvec_dcable_augwithzeros/checkpoints"
#path = f"./trained_models/all_mb_03_27/new_mb_03_27_poc64_lr5em4_bs128_sep_diff_rotvec_cable_augwithzeros/checkpoints"
#path = f"./trained_models/all_mb_03_27/new_mb_03_27_poc64_lr5em4_bs128_sep_nodiff_rotmat_dcable_noaug/checkpoints"
#path = f"./trained_models/all_mb_03_27/new_mb_03_27_poc64_lr5em4_bs128_sep_nodiff_rotmat_dcable_augwithzeros/checkpoints"
#path = f"./trained_models/all_mb_03_27/new_mb_03_27_poc64_lr5em4_bs128_inbilstm_nodiff_rotvec_cable_noaug/checkpoints"
#path = f"./trained_models/all_mb_03_27/new_mb_03_27_poc64_lr5em4_bs128_inbilstm_nodiff_rotvec_cable_augwithzeros/checkpoints"
path = f"./trained_models/all_mb_zoval_04_25/new_mb_zoval_04_25_poc64_lr5em4_bs128_sep_nodiff_rotmat_dcable_augwithzeros/checkpoints"
name = path.split("/")[-2]
name_fields = name.split("_")
m = name_fields[-5]
d = name_fields[-4]
q = name_fields[-3]
c = name_fields[-2]
a = name_fields[-1]
class args:
batch_size = 128
working_dir = './trainings'
#dataset_path = f"./data/prepared_datasets/new_mb_03_27_poc64/train.tsv"
dataset_path = f"./data/prepared_datasets/new_mb_zoval_04_25/train.tsv"
train_ds, train_size, tX1, tX2, tX3, tY = prepare_dataset_cond(args.dataset_path, rot=q, diff=(d == "diff"), augment=(a == "augwithzeros"))
val_ds, val_size, vX1, vX2, vX3, vY = prepare_dataset_cond(args.dataset_path.replace("train", "val"), rot=q, diff=(d == "diff"))
test_ds, test_size, teX1, teX2, teX3, teY = prepare_dataset_cond(args.dataset_path.replace("train", "test"), rot=q, diff=(d == "diff"))
ds_stats = compute_ds_stats(train_ds)
#ds, ds_size = train_ds, train_size
#ds, ds_size = val_ds, val_size
ds, ds_size = test_ds, test_size
loss = CablePtsLoss()
if m == "sep":
model = SeparatedNeuralPredictor()
elif m == "cnn":
model = CNN()
elif m == "inbilstm":
model = INBiLSTM()
else:
print("WRONG MODEL NAME")
assert False
ckpt = tf.train.Checkpoint(model=model)
best_list = list(glob(os.path.join(path, "best-*.index")))
assert best_list
best = best_list[0][:-6]
#ckpt.restore(best).expect_partial()
#ckpt.restore(best)
def inference(rotation, translation, cable):
rotation_, translation_, cable_ = whitening(rotation, translation, cable, ds_stats)
R_l_0, R_l_1, R_r_0, R_r_1 = unpack_rotation(rotation_)
t_l_0, t_l_1 = unpack_translation(translation_)
cable_, dcable_ = unpack_cable(cable_)
y_pred_ = model((R_l_0, R_l_1, R_r_0, R_r_1), (t_l_0, t_l_1), dcable_ if c == "dcable" else cable_,
unpack_rotation(rotation), unpack_translation(translation))
# y_pred_ = model((R_l_0, R_l_1, R_r_0, R_r_1), (t_l_0, t_l_1), dcable_ if ifdcable else cable_)
# y_pred = y_pred_ * ds_stats["sy"] + ds_stats["my"]
y_pred = y_pred_
return y_pred + cable[..., :3]
dataset_epoch = ds.shuffle(ds_size)
dataset_epoch = dataset_epoch.batch(args.batch_size).prefetch(args.batch_size)
epoch_loss = []
pts_losses_abs = []
pts_losses_euc = []
ratio_losses = []
for i, rotation, translation, cable, y_gt in _ds('Train', dataset_epoch, ds_size, 0, args.batch_size):
y_pred = inference(rotation, translation, cable)
pts_loss_abs, pts_loss_euc, pts_loss_l2 = loss(y_gt, y_pred)
cable, dcable_ = unpack_cable(cable)
for i in range(y_pred.shape[0]):
# frechet_dist_gtpred = frdist(y_gt[i], y_pred[i])
# frechet_dist_gtcable = frdist(y_gt[i], cable[i])
# dtw_gtpred = dtw(y_gt[i], y_pred[i])
# dtw_gtcable = dtw(y_gt[i], cable[i])
L3_gtpred = calculateL3(y_gt[i].numpy().T, y_pred[i].numpy().T)
L3_gtcable = calculateL3(y_gt[i].numpy().T, cable[i].numpy().T)
ratio_loss = L3_gtpred / (L3_gtcable + 1e-8)
print("L3:", L3_gtpred)
print("RATIO:", ratio_loss)
ratio_losses.append(ratio_loss)
pts_losses_abs.append(pts_loss_abs)
pts_losses_euc.append(pts_loss_euc)
pts_loss_abs = tf.concat(pts_losses_abs, -1).numpy()
pts_loss_euc = tf.concat(pts_losses_euc, -1).numpy()
ratio_losses = np.array(ratio_losses)
mean_pts_losses_abs = np.mean(pts_loss_abs)
mean_pts_losses_euc = np.mean(pts_loss_euc)
mean_ratio_losses = np.mean(ratio_losses)
std_pts_losses_abs = np.std(pts_loss_abs)
std_pts_losses_euc = np.std(pts_loss_euc)
std_ratio_losses = np.std(ratio_losses)
#print("EPOCH LOSS:", epoch_loss)
#print("PREDICTION LOSS:", prediction_losses)
#print("CP ABS LOSS:", cp_losses_abs)
#print("PTS EUC LOSS:", pts_losses_euc)
results = {
# "mean_ratio_loss": mean_ratio_losses, "std_ratio_loss": std_ratio_losses,
# "mean_pts_loss_abs": mean_pts_losses_abs, "std_pts_loss_abs": std_pts_losses_abs,
# "mean_pts_loss_euc": mean_pts_losses_euc, "std_pts_loss_euc": std_pts_losses_euc,
"ratio_loss": ratio_losses,
"pts_loss_abs": pts_losses_abs,
"pts_loss_euc": pts_losses_euc,
}
#np.save(f"results/all_mb_03_27/{name}.npy", results)
os.makedirs("results/all_mb_zoval_04_25/", exist_ok=True)
np.save(f"results/all_mb_zoval_04_25/{name}.npy", results)